Moral Policing

Establishing a Cause & Effect Relationship

Does X lead to Y i.e. does X cause the effect Y? Well it might, but how do you validate this relationship?

A lot of thought and effort goes into designing such an argument. It is not always possible to design perfect cause & effect (C&E) relationships in business because of its dynamic nature, but we can always question a C&E relationship to validate it to a reasonable extent. Validated C&E relationships lead to better decision making, however, validation comes at a cost.

Lets take an example to show how a sound argument is designed (example courtesy Prof Amit Das at NBS)

Suppose we claim that a particular training program improves productivity.

How do we support such a claim? To start with we could show that participants coming out of the program are productive.

But someone could challenge us saying that those who participated in the training are no more productive than those who did not. To counter this, we will need to measure the productivity of both, the people who attended the training and people who did not and show that there indeed is a difference in the productivity levels of the two groups.

However, someone might again challenge us saying that the people who participated in the training were already more productive than others (i.e. training did not increase productivity rather the people chosen were already more productive). We can dismiss this argument by measuring the productivity of the training participants before and after the training and show that there was indeed an increase in the productivity.

Next, we could be challenged with the argument that people who did not attend the training also became more productive over time (i.e. everyone gets more productive over time and training did not have an effect on productivity). To thwart this argument, we can measure the productivity (before and after) for the people who attended the training as well as for the people who did not. We can then show that the people who attended the training had a relatively higher increase in productivity levels than the people who did not attend the training.

The challenger's next argument could be that the people who were sent to the training were smarter than the people who were not so they improved faster than the rest (i.e. the increase in productivity was not on account of training). To eliminate this, we have to make sure that those who went for training and those who did not were comparable people.

Another argument against us could be that the productivity improvement was not because of training as such but maybe due to other differences like training location (maybe Bahamas) which might have improved the morale of the participants and hence led to productivity improvements. Well, to dismiss this argument we could have two groups attend two different training programs in the same place and then measure the productivity of the two groups and show that our training program is more effective, other things kept constant. (Frankly speaking this is going too far, most probably we would not have to go to this extent for business arguments).

All said and done, some thought put in designing our argument goes a long way in determining how useful will that argument be for decision making. A casually made argument without proper validations might lead to erroneous decisions costing companies pots of money (Say if in our case the training program was actually not useful, but just a casual argument like 'the training increased productivity' might cause a company to spend loads on that training).

Lessons to be learnt -
1. If designing an argument, put in a little thought and make it as sound as possible.
2. If encountered with an argument, do not accept it without challenging its validity.

Wrong Metrics and Relationships - The reason behind Business Inefficiencies?

" The salesman who achieves his sales targets should be rewarded"
" Installing SAP helped my department reduce costs by 20%"
" Company X is better than company Y because it has more revenues and its sales grew by 50%"

Familiar sounding statements? The next time you see such a statement, think about it for a second. Metrics/Statistics are as much wrong as they are right. Many times, statisticians only give you the good information and hide the things which you should not see. Many times we ourselves try and link two things which might actually be totally unrelated or might not have a strong correlation.

Many businesses have a "sales achieved" metric for their salesmen. Salesman 'A' might be able to sell 100 units but with a margin of only $1 on each unit; While Salesman 'B' might be able to sell 50 units with a margin of $2 on each unit. Both of them made $100 for the company, but only salesman A was able to achieve the target of 100 units sale. What the managers have wrongly assumed here is "Sales target achieved => better salesman"

There are loads and loads of examples where such a wrongly established relationship leads to a lot of inefficiencies in the business. With such a metric in place, the objective of the salesman becomes to sell more and not for how much margin. The objective of the company is to make more money but the objective of the salesman is not to make money but to sell more...all because of a wrong metric/performance measure.

Incorrect 'cause and effect' relationships have rocked many a business. Many firms are gungho on Technology these days. People claim that a new software reduced costs by 20% for their department. Yeah ok, it must have...but why do you not tell us about the increase in overhead costs (which may not be allocated to that department) due to that software. So the costs for department might have reduced, but the costs for the company might not have!

Take the example of the Fortune 500 rankings. The rankings are done on the basis of gross revenues. The companies which make it to the ranking proudly claim that they are a Fortune 500 company and we run behind those companies for jobs and even metion in our resumes "Worked for a Fortune 500 client". Do you think gross revenues is the correct measure? Tomorrow, I could open a company, make 10 billion laptops and sell them for $10 each. This would give me revenues of $100 billion and would make my company rank #18 in the Fortune 500 rankings. Yeah, I got the revenues but made a huge loss! Profitability (profit as a percent of revenues) would any day be a better metric to give us a more realistic picture of which company is performing well.

Statisticians are not the only people to be blamed, normal people like you and me are to blame as much. We try and make frameworks and models out of everything we see and try and correlate things which in reality might not be correlated. We see X leading to Y and make a notion in our head that X leads to Y. Does it, really? It might, it might not.

Check the next post to see how to establish a cause and effect relationship.

Democratic values?

Came across this recently -

"In USA you can kiss in public places but cannot shit; in India you can shit in public places but cannot kiss"

Very aptly describes how democracy could mean different things in different places.